Pi v0.79.0 adds project trust gating

// 1h agoPRODUCT UPDATE

Pi v0.79.0 adds project trust gating

ANNOUNCEMENT PRODUCT GITHUB PRODUCT HUNT X

Pi v0.79.0 introduces project trust gating to prompt users before loading potentially untrusted project-local settings, instructions, and packages. The update also corrects TUI rendering, autocomplete behavior, prompt history navigation, and routing issues on OpenAI-compatible providers.

// ANALYSIS

The addition of project trust gating shows that agent toolkits are maturing to address security risks associated with running autonomous tools on untrusted code repositories, a critical step for enterprise and professional adoption.

–**Security Maturity**: Restricting automatic execution of repository-level settings prevents potential prompt injection or malicious code execution via local configuration files.
–**Provider Reliability**: Fixing routing preferences for custom providers ensures developers can reliably use OpenRouter and OpenAI-compatible endpoints.
–**UX Polish**: Key TUI fixes resolve annoying UI glitches like stale lines and cursor position jumps in autocomplete and history navigation.

// TAGS

agentdevtoolopen-sourceclicybersecurity

DISCOVERED

1h ago

2026-06-08

PUBLISHED

2h ago

2026-06-08

RELEVANCE

7/ 10

AUTHOR

PiChangelog

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

OPEN SOURCE35m ago

llama.cpp merges Gemma 4 MTP support

llama.cpp has officially merged support for Gemma 4 Multi-Token Prediction (MTP), enabling developers to leverage speculative decoding techniques directly on local hardware. By pairing Gemma 4 MTP with Gemma 4 Quantization Aware Training (QAT), developers can create fast, lightweight setups that deliver high-speed inference without the overhead of cloud hosting.

BENCHMARK37m ago

Hugging Face, Mecado release CADGenBench

Hugging Face and Mecado have released CADGenBench, an open-source, tool-agnostic benchmark to evaluate AI systems on generating and editing engineering-grade 3D mechanical parts. Submissions are scored via standard STEP files on a Hugging Face Space against private ground truth data.

BENCHMARK37m ago

Grok Imagine Video 1.5 Preview tops Design Arena

Grok Imagine Video 1.5 Preview has officially claimed the number-one ranking in the Image-to-Video generation category on the Design Arena benchmark platform. This achievement highlights the model's competitive performance in crowdsourced human-evaluated design benchmarks, demonstrating xAI's growing capabilities in generative media.